EMO-BOOST: Emotion-Augmented Audio-Visual Features for Improved Generalization in Deepfake Detection

In three linesEMO-BOOST combines RGB/audio-focused deepfake detectors with EmoForensics, an emotion-based detector using audio-visual emotion recognition. The method models temporal consistency of emotions and improves cross-manipulation generalization by 2.1% AUC on FakeAVCeleb.

Summary generated by Claude — human-verified